1 /* 2 * Copyright (C) 2005 Anthony Liguori <anthony@codemonkey.ws> 3 * 4 * Network Block Device 5 * 6 * This program is free software; you can redistribute it and/or modify 7 * it under the terms of the GNU General Public License as published by 8 * the Free Software Foundation; under version 2 of the License. 9 * 10 * This program is distributed in the hope that it will be useful, 11 * but WITHOUT ANY WARRANTY; without even the implied warranty of 12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 13 * GNU General Public License for more details. 14 * 15 * You should have received a copy of the GNU General Public License 16 * along with this program; if not, see <http://www.gnu.org/licenses/>. 17 */ 18 19 #include "qemu/osdep.h" 20 #include <getopt.h> 21 #include <libgen.h> 22 #include <pthread.h> 23 24 #include "qemu-common.h" 25 #include "qapi/error.h" 26 #include "qemu/cutils.h" 27 #include "sysemu/block-backend.h" 28 #include "block/block_int.h" 29 #include "block/nbd.h" 30 #include "qemu/main-loop.h" 31 #include "qemu/option.h" 32 #include "qemu/error-report.h" 33 #include "qemu/config-file.h" 34 #include "qemu/bswap.h" 35 #include "qemu/log.h" 36 #include "qemu/systemd.h" 37 #include "block/snapshot.h" 38 #include "qapi/qmp/qdict.h" 39 #include "qapi/qmp/qstring.h" 40 #include "qom/object_interfaces.h" 41 #include "io/channel-socket.h" 42 #include "io/net-listener.h" 43 #include "crypto/init.h" 44 #include "trace/control.h" 45 #include "qemu-version.h" 46 47 #ifdef __linux__ 48 #define HAVE_NBD_DEVICE 1 49 #else 50 #define HAVE_NBD_DEVICE 0 51 #endif 52 53 #define SOCKET_PATH "/var/lock/qemu-nbd-%s" 54 #define QEMU_NBD_OPT_CACHE 256 55 #define QEMU_NBD_OPT_AIO 257 56 #define QEMU_NBD_OPT_DISCARD 258 57 #define QEMU_NBD_OPT_DETECT_ZEROES 259 58 #define QEMU_NBD_OPT_OBJECT 260 59 #define QEMU_NBD_OPT_TLSCREDS 261 60 #define QEMU_NBD_OPT_IMAGE_OPTS 262 61 #define QEMU_NBD_OPT_FORK 263 62 #define QEMU_NBD_OPT_TLSAUTHZ 264 63 64 #define MBR_SIZE 512 65 66 static NBDExport *export; 67 static int verbose; 68 static char *srcpath; 69 static SocketAddress *saddr; 70 static int persistent = 0; 71 static enum { RUNNING, TERMINATE, TERMINATING, TERMINATED } state; 72 static int shared = 1; 73 static int nb_fds; 74 static QIONetListener *server; 75 static QCryptoTLSCreds *tlscreds; 76 static const char *tlsauthz; 77 78 static void usage(const char *name) 79 { 80 (printf) ( 81 "Usage: %s [OPTIONS] FILE\n" 82 " or: %s -L [OPTIONS]\n" 83 "QEMU Disk Network Block Device Utility\n" 84 "\n" 85 " -h, --help display this help and exit\n" 86 " -V, --version output version information and exit\n" 87 "\n" 88 "Connection properties:\n" 89 " -p, --port=PORT port to listen on (default `%d')\n" 90 " -b, --bind=IFACE interface to bind to (default `0.0.0.0')\n" 91 " -k, --socket=PATH path to the unix socket\n" 92 " (default '"SOCKET_PATH"')\n" 93 " -e, --shared=NUM device can be shared by NUM clients (default '1')\n" 94 " -t, --persistent don't exit on the last connection\n" 95 " -v, --verbose display extra debugging information\n" 96 " -x, --export-name=NAME expose export by name (default is empty string)\n" 97 " -D, --description=TEXT export a human-readable description\n" 98 "\n" 99 "Exposing part of the image:\n" 100 " -o, --offset=OFFSET offset into the image\n" 101 " -P, --partition=NUM only expose partition NUM\n" 102 " -B, --bitmap=NAME expose a persistent dirty bitmap\n" 103 "\n" 104 "General purpose options:\n" 105 " -L, --list list exports available from another NBD server\n" 106 " --object type,id=ID,... define an object such as 'secret' for providing\n" 107 " passwords and/or encryption keys\n" 108 " --tls-creds=ID use id of an earlier --object to provide TLS\n" 109 " --tls-authz=ID use id of an earlier --object to provide\n" 110 " authorization\n" 111 " -T, --trace [[enable=]<pattern>][,events=<file>][,file=<file>]\n" 112 " specify tracing options\n" 113 " --fork fork off the server process and exit the parent\n" 114 " once the server is running\n" 115 #if HAVE_NBD_DEVICE 116 "\n" 117 "Kernel NBD client support:\n" 118 " -c, --connect=DEV connect FILE to the local NBD device DEV\n" 119 " -d, --disconnect disconnect the specified device\n" 120 #endif 121 "\n" 122 "Block device options:\n" 123 " -f, --format=FORMAT set image format (raw, qcow2, ...)\n" 124 " -r, --read-only export read-only\n" 125 " -s, --snapshot use FILE as an external snapshot, create a temporary\n" 126 " file with backing_file=FILE, redirect the write to\n" 127 " the temporary one\n" 128 " -l, --load-snapshot=SNAPSHOT_PARAM\n" 129 " load an internal snapshot inside FILE and export it\n" 130 " as an read-only device, SNAPSHOT_PARAM format is\n" 131 " 'snapshot.id=[ID],snapshot.name=[NAME]', or\n" 132 " '[ID_OR_NAME]'\n" 133 " -n, --nocache disable host cache\n" 134 " --cache=MODE set cache mode (none, writeback, ...)\n" 135 " --aio=MODE set AIO mode (native or threads)\n" 136 " --discard=MODE set discard mode (ignore, unmap)\n" 137 " --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n" 138 " --image-opts treat FILE as a full set of image options\n" 139 "\n" 140 QEMU_HELP_BOTTOM "\n" 141 , name, name, NBD_DEFAULT_PORT, "DEVICE"); 142 } 143 144 static void version(const char *name) 145 { 146 printf( 147 "%s " QEMU_FULL_VERSION "\n" 148 "Written by Anthony Liguori.\n" 149 "\n" 150 QEMU_COPYRIGHT "\n" 151 "This is free software; see the source for copying conditions. There is NO\n" 152 "warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.\n" 153 , name); 154 } 155 156 struct partition_record 157 { 158 uint8_t bootable; 159 uint8_t start_head; 160 uint32_t start_cylinder; 161 uint8_t start_sector; 162 uint8_t system; 163 uint8_t end_head; 164 uint8_t end_cylinder; 165 uint8_t end_sector; 166 uint32_t start_sector_abs; 167 uint32_t nb_sectors_abs; 168 }; 169 170 static void read_partition(uint8_t *p, struct partition_record *r) 171 { 172 r->bootable = p[0]; 173 r->start_head = p[1]; 174 r->start_cylinder = p[3] | ((p[2] << 2) & 0x0300); 175 r->start_sector = p[2] & 0x3f; 176 r->system = p[4]; 177 r->end_head = p[5]; 178 r->end_cylinder = p[7] | ((p[6] << 2) & 0x300); 179 r->end_sector = p[6] & 0x3f; 180 181 r->start_sector_abs = ldl_le_p(p + 8); 182 r->nb_sectors_abs = ldl_le_p(p + 12); 183 } 184 185 static int find_partition(BlockBackend *blk, int partition, 186 uint64_t *offset, uint64_t *size) 187 { 188 struct partition_record mbr[4]; 189 uint8_t data[MBR_SIZE]; 190 int i; 191 int ext_partnum = 4; 192 int ret; 193 194 ret = blk_pread(blk, 0, data, sizeof(data)); 195 if (ret < 0) { 196 error_report("error while reading: %s", strerror(-ret)); 197 exit(EXIT_FAILURE); 198 } 199 200 if (data[510] != 0x55 || data[511] != 0xaa) { 201 return -EINVAL; 202 } 203 204 for (i = 0; i < 4; i++) { 205 read_partition(&data[446 + 16 * i], &mbr[i]); 206 207 if (!mbr[i].system || !mbr[i].nb_sectors_abs) { 208 continue; 209 } 210 211 if (mbr[i].system == 0xF || mbr[i].system == 0x5) { 212 struct partition_record ext[4]; 213 uint8_t data1[MBR_SIZE]; 214 int j; 215 216 ret = blk_pread(blk, mbr[i].start_sector_abs * MBR_SIZE, 217 data1, sizeof(data1)); 218 if (ret < 0) { 219 error_report("error while reading: %s", strerror(-ret)); 220 exit(EXIT_FAILURE); 221 } 222 223 for (j = 0; j < 4; j++) { 224 read_partition(&data1[446 + 16 * j], &ext[j]); 225 if (!ext[j].system || !ext[j].nb_sectors_abs) { 226 continue; 227 } 228 229 if ((ext_partnum + j + 1) == partition) { 230 *offset = (uint64_t)ext[j].start_sector_abs << 9; 231 *size = (uint64_t)ext[j].nb_sectors_abs << 9; 232 return 0; 233 } 234 } 235 ext_partnum += 4; 236 } else if ((i + 1) == partition) { 237 *offset = (uint64_t)mbr[i].start_sector_abs << 9; 238 *size = (uint64_t)mbr[i].nb_sectors_abs << 9; 239 return 0; 240 } 241 } 242 243 return -ENOENT; 244 } 245 246 static void termsig_handler(int signum) 247 { 248 atomic_cmpxchg(&state, RUNNING, TERMINATE); 249 qemu_notify_event(); 250 } 251 252 253 static int qemu_nbd_client_list(SocketAddress *saddr, QCryptoTLSCreds *tls, 254 const char *hostname) 255 { 256 int ret = EXIT_FAILURE; 257 int rc; 258 Error *err = NULL; 259 QIOChannelSocket *sioc; 260 NBDExportInfo *list; 261 int i, j; 262 263 sioc = qio_channel_socket_new(); 264 if (qio_channel_socket_connect_sync(sioc, saddr, &err) < 0) { 265 error_report_err(err); 266 return EXIT_FAILURE; 267 } 268 rc = nbd_receive_export_list(QIO_CHANNEL(sioc), tls, hostname, &list, 269 &err); 270 if (rc < 0) { 271 if (err) { 272 error_report_err(err); 273 } 274 goto out; 275 } 276 printf("exports available: %d\n", rc); 277 for (i = 0; i < rc; i++) { 278 printf(" export: '%s'\n", list[i].name); 279 if (list[i].description && *list[i].description) { 280 printf(" description: %s\n", list[i].description); 281 } 282 if (list[i].flags & NBD_FLAG_HAS_FLAGS) { 283 static const char *const flag_names[] = { 284 [NBD_FLAG_READ_ONLY_BIT] = "readonly", 285 [NBD_FLAG_SEND_FLUSH_BIT] = "flush", 286 [NBD_FLAG_SEND_FUA_BIT] = "fua", 287 [NBD_FLAG_ROTATIONAL_BIT] = "rotational", 288 [NBD_FLAG_SEND_TRIM_BIT] = "trim", 289 [NBD_FLAG_SEND_WRITE_ZEROES_BIT] = "zeroes", 290 [NBD_FLAG_SEND_DF_BIT] = "df", 291 [NBD_FLAG_CAN_MULTI_CONN_BIT] = "multi", 292 [NBD_FLAG_SEND_RESIZE_BIT] = "resize", 293 [NBD_FLAG_SEND_CACHE_BIT] = "cache", 294 }; 295 296 printf(" size: %" PRIu64 "\n", list[i].size); 297 printf(" flags: 0x%x (", list[i].flags); 298 for (size_t bit = 0; bit < ARRAY_SIZE(flag_names); bit++) { 299 if (flag_names[bit] && (list[i].flags & (1 << bit))) { 300 printf(" %s", flag_names[bit]); 301 } 302 } 303 printf(" )\n"); 304 } 305 if (list[i].min_block) { 306 printf(" min block: %u\n", list[i].min_block); 307 printf(" opt block: %u\n", list[i].opt_block); 308 printf(" max block: %u\n", list[i].max_block); 309 } 310 if (list[i].n_contexts) { 311 printf(" available meta contexts: %d\n", list[i].n_contexts); 312 for (j = 0; j < list[i].n_contexts; j++) { 313 printf(" %s\n", list[i].contexts[j]); 314 } 315 } 316 } 317 nbd_free_export_list(list, rc); 318 319 ret = EXIT_SUCCESS; 320 out: 321 object_unref(OBJECT(sioc)); 322 return ret; 323 } 324 325 326 #if HAVE_NBD_DEVICE 327 static void *show_parts(void *arg) 328 { 329 char *device = arg; 330 int nbd; 331 332 /* linux just needs an open() to trigger 333 * the partition table update 334 * but remember to load the module with max_part != 0 : 335 * modprobe nbd max_part=63 336 */ 337 nbd = open(device, O_RDWR); 338 if (nbd >= 0) { 339 close(nbd); 340 } 341 return NULL; 342 } 343 344 static void *nbd_client_thread(void *arg) 345 { 346 char *device = arg; 347 NBDExportInfo info = { .request_sizes = false, .name = g_strdup("") }; 348 QIOChannelSocket *sioc; 349 int fd; 350 int ret; 351 pthread_t show_parts_thread; 352 Error *local_error = NULL; 353 354 sioc = qio_channel_socket_new(); 355 if (qio_channel_socket_connect_sync(sioc, 356 saddr, 357 &local_error) < 0) { 358 error_report_err(local_error); 359 goto out; 360 } 361 362 ret = nbd_receive_negotiate(QIO_CHANNEL(sioc), 363 NULL, NULL, NULL, &info, &local_error); 364 if (ret < 0) { 365 if (local_error) { 366 error_report_err(local_error); 367 } 368 goto out_socket; 369 } 370 371 fd = open(device, O_RDWR); 372 if (fd < 0) { 373 /* Linux-only, we can use %m in printf. */ 374 error_report("Failed to open %s: %m", device); 375 goto out_socket; 376 } 377 378 ret = nbd_init(fd, sioc, &info, &local_error); 379 if (ret < 0) { 380 error_report_err(local_error); 381 goto out_fd; 382 } 383 384 /* update partition table */ 385 pthread_create(&show_parts_thread, NULL, show_parts, device); 386 387 if (verbose) { 388 fprintf(stderr, "NBD device %s is now connected to %s\n", 389 device, srcpath); 390 } else { 391 /* Close stderr so that the qemu-nbd process exits. */ 392 dup2(STDOUT_FILENO, STDERR_FILENO); 393 } 394 395 ret = nbd_client(fd); 396 if (ret) { 397 goto out_fd; 398 } 399 close(fd); 400 object_unref(OBJECT(sioc)); 401 g_free(info.name); 402 kill(getpid(), SIGTERM); 403 return (void *) EXIT_SUCCESS; 404 405 out_fd: 406 close(fd); 407 out_socket: 408 object_unref(OBJECT(sioc)); 409 out: 410 g_free(info.name); 411 kill(getpid(), SIGTERM); 412 return (void *) EXIT_FAILURE; 413 } 414 #endif /* HAVE_NBD_DEVICE */ 415 416 static int nbd_can_accept(void) 417 { 418 return state == RUNNING && nb_fds < shared; 419 } 420 421 static void nbd_export_closed(NBDExport *export) 422 { 423 assert(state == TERMINATING); 424 state = TERMINATED; 425 } 426 427 static void nbd_update_server_watch(void); 428 429 static void nbd_client_closed(NBDClient *client, bool negotiated) 430 { 431 nb_fds--; 432 if (negotiated && nb_fds == 0 && !persistent && state == RUNNING) { 433 state = TERMINATE; 434 } 435 nbd_update_server_watch(); 436 nbd_client_put(client); 437 } 438 439 static void nbd_accept(QIONetListener *listener, QIOChannelSocket *cioc, 440 gpointer opaque) 441 { 442 if (state >= TERMINATE) { 443 return; 444 } 445 446 nb_fds++; 447 nbd_update_server_watch(); 448 nbd_client_new(cioc, tlscreds, tlsauthz, nbd_client_closed); 449 } 450 451 static void nbd_update_server_watch(void) 452 { 453 if (nbd_can_accept()) { 454 qio_net_listener_set_client_func(server, nbd_accept, NULL, NULL); 455 } else { 456 qio_net_listener_set_client_func(server, NULL, NULL, NULL); 457 } 458 } 459 460 461 static SocketAddress *nbd_build_socket_address(const char *sockpath, 462 const char *bindto, 463 const char *port) 464 { 465 SocketAddress *saddr; 466 467 saddr = g_new0(SocketAddress, 1); 468 if (sockpath) { 469 saddr->type = SOCKET_ADDRESS_TYPE_UNIX; 470 saddr->u.q_unix.path = g_strdup(sockpath); 471 } else { 472 InetSocketAddress *inet; 473 saddr->type = SOCKET_ADDRESS_TYPE_INET; 474 inet = &saddr->u.inet; 475 inet->host = g_strdup(bindto); 476 if (port) { 477 inet->port = g_strdup(port); 478 } else { 479 inet->port = g_strdup_printf("%d", NBD_DEFAULT_PORT); 480 } 481 } 482 483 return saddr; 484 } 485 486 487 static QemuOptsList file_opts = { 488 .name = "file", 489 .implied_opt_name = "file", 490 .head = QTAILQ_HEAD_INITIALIZER(file_opts.head), 491 .desc = { 492 /* no elements => accept any params */ 493 { /* end of list */ } 494 }, 495 }; 496 497 static QemuOptsList qemu_object_opts = { 498 .name = "object", 499 .implied_opt_name = "qom-type", 500 .head = QTAILQ_HEAD_INITIALIZER(qemu_object_opts.head), 501 .desc = { 502 { } 503 }, 504 }; 505 506 507 508 static QCryptoTLSCreds *nbd_get_tls_creds(const char *id, bool list, 509 Error **errp) 510 { 511 Object *obj; 512 QCryptoTLSCreds *creds; 513 514 obj = object_resolve_path_component( 515 object_get_objects_root(), id); 516 if (!obj) { 517 error_setg(errp, "No TLS credentials with id '%s'", 518 id); 519 return NULL; 520 } 521 creds = (QCryptoTLSCreds *) 522 object_dynamic_cast(obj, TYPE_QCRYPTO_TLS_CREDS); 523 if (!creds) { 524 error_setg(errp, "Object with id '%s' is not TLS credentials", 525 id); 526 return NULL; 527 } 528 529 if (list) { 530 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_CLIENT) { 531 error_setg(errp, 532 "Expecting TLS credentials with a client endpoint"); 533 return NULL; 534 } 535 } else { 536 if (creds->endpoint != QCRYPTO_TLS_CREDS_ENDPOINT_SERVER) { 537 error_setg(errp, 538 "Expecting TLS credentials with a server endpoint"); 539 return NULL; 540 } 541 } 542 object_ref(obj); 543 return creds; 544 } 545 546 static void setup_address_and_port(const char **address, const char **port) 547 { 548 if (*address == NULL) { 549 *address = "0.0.0.0"; 550 } 551 552 if (*port == NULL) { 553 *port = stringify(NBD_DEFAULT_PORT); 554 } 555 } 556 557 /* 558 * Check socket parameters compatibility when socket activation is used. 559 */ 560 static const char *socket_activation_validate_opts(const char *device, 561 const char *sockpath, 562 const char *address, 563 const char *port, 564 bool list) 565 { 566 if (device != NULL) { 567 return "NBD device can't be set when using socket activation"; 568 } 569 570 if (sockpath != NULL) { 571 return "Unix socket can't be set when using socket activation"; 572 } 573 574 if (address != NULL) { 575 return "The interface can't be set when using socket activation"; 576 } 577 578 if (port != NULL) { 579 return "TCP port number can't be set when using socket activation"; 580 } 581 582 if (list) { 583 return "List mode is incompatible with socket activation"; 584 } 585 586 return NULL; 587 } 588 589 static void qemu_nbd_shutdown(void) 590 { 591 job_cancel_sync_all(); 592 bdrv_close_all(); 593 } 594 595 int main(int argc, char **argv) 596 { 597 BlockBackend *blk; 598 BlockDriverState *bs; 599 uint64_t dev_offset = 0; 600 uint16_t nbdflags = 0; 601 bool disconnect = false; 602 const char *bindto = NULL; 603 const char *port = NULL; 604 char *sockpath = NULL; 605 char *device = NULL; 606 int64_t fd_size; 607 QemuOpts *sn_opts = NULL; 608 const char *sn_id_or_name = NULL; 609 const char *sopt = "hVb:o:p:rsnP:c:dvk:e:f:tl:x:T:D:B:L"; 610 struct option lopt[] = { 611 { "help", no_argument, NULL, 'h' }, 612 { "version", no_argument, NULL, 'V' }, 613 { "bind", required_argument, NULL, 'b' }, 614 { "port", required_argument, NULL, 'p' }, 615 { "socket", required_argument, NULL, 'k' }, 616 { "offset", required_argument, NULL, 'o' }, 617 { "read-only", no_argument, NULL, 'r' }, 618 { "partition", required_argument, NULL, 'P' }, 619 { "bitmap", required_argument, NULL, 'B' }, 620 { "connect", required_argument, NULL, 'c' }, 621 { "disconnect", no_argument, NULL, 'd' }, 622 { "list", no_argument, NULL, 'L' }, 623 { "snapshot", no_argument, NULL, 's' }, 624 { "load-snapshot", required_argument, NULL, 'l' }, 625 { "nocache", no_argument, NULL, 'n' }, 626 { "cache", required_argument, NULL, QEMU_NBD_OPT_CACHE }, 627 { "aio", required_argument, NULL, QEMU_NBD_OPT_AIO }, 628 { "discard", required_argument, NULL, QEMU_NBD_OPT_DISCARD }, 629 { "detect-zeroes", required_argument, NULL, 630 QEMU_NBD_OPT_DETECT_ZEROES }, 631 { "shared", required_argument, NULL, 'e' }, 632 { "format", required_argument, NULL, 'f' }, 633 { "persistent", no_argument, NULL, 't' }, 634 { "verbose", no_argument, NULL, 'v' }, 635 { "object", required_argument, NULL, QEMU_NBD_OPT_OBJECT }, 636 { "export-name", required_argument, NULL, 'x' }, 637 { "description", required_argument, NULL, 'D' }, 638 { "tls-creds", required_argument, NULL, QEMU_NBD_OPT_TLSCREDS }, 639 { "tls-authz", required_argument, NULL, QEMU_NBD_OPT_TLSAUTHZ }, 640 { "image-opts", no_argument, NULL, QEMU_NBD_OPT_IMAGE_OPTS }, 641 { "trace", required_argument, NULL, 'T' }, 642 { "fork", no_argument, NULL, QEMU_NBD_OPT_FORK }, 643 { NULL, 0, NULL, 0 } 644 }; 645 int ch; 646 int opt_ind = 0; 647 int flags = BDRV_O_RDWR; 648 int partition = 0; 649 int ret = 0; 650 bool seen_cache = false; 651 bool seen_discard = false; 652 bool seen_aio = false; 653 pthread_t client_thread; 654 const char *fmt = NULL; 655 Error *local_err = NULL; 656 BlockdevDetectZeroesOptions detect_zeroes = BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF; 657 QDict *options = NULL; 658 const char *export_name = NULL; /* defaults to "" later for server mode */ 659 const char *export_description = NULL; 660 const char *bitmap = NULL; 661 const char *tlscredsid = NULL; 662 bool imageOpts = false; 663 bool writethrough = true; 664 char *trace_file = NULL; 665 bool fork_process = false; 666 bool list = false; 667 int old_stderr = -1; 668 unsigned socket_activation; 669 670 /* The client thread uses SIGTERM to interrupt the server. A signal 671 * handler ensures that "qemu-nbd -v -c" exits with a nice status code. 672 */ 673 struct sigaction sa_sigterm; 674 memset(&sa_sigterm, 0, sizeof(sa_sigterm)); 675 sa_sigterm.sa_handler = termsig_handler; 676 sigaction(SIGTERM, &sa_sigterm, NULL); 677 678 #ifdef CONFIG_POSIX 679 signal(SIGPIPE, SIG_IGN); 680 #endif 681 682 error_init(argv[0]); 683 module_call_init(MODULE_INIT_TRACE); 684 qcrypto_init(&error_fatal); 685 686 module_call_init(MODULE_INIT_QOM); 687 qemu_add_opts(&qemu_object_opts); 688 qemu_add_opts(&qemu_trace_opts); 689 qemu_init_exec_dir(argv[0]); 690 691 while ((ch = getopt_long(argc, argv, sopt, lopt, &opt_ind)) != -1) { 692 switch (ch) { 693 case 's': 694 flags |= BDRV_O_SNAPSHOT; 695 break; 696 case 'n': 697 optarg = (char *) "none"; 698 /* fallthrough */ 699 case QEMU_NBD_OPT_CACHE: 700 if (seen_cache) { 701 error_report("-n and --cache can only be specified once"); 702 exit(EXIT_FAILURE); 703 } 704 seen_cache = true; 705 if (bdrv_parse_cache_mode(optarg, &flags, &writethrough) == -1) { 706 error_report("Invalid cache mode `%s'", optarg); 707 exit(EXIT_FAILURE); 708 } 709 break; 710 case QEMU_NBD_OPT_AIO: 711 if (seen_aio) { 712 error_report("--aio can only be specified once"); 713 exit(EXIT_FAILURE); 714 } 715 seen_aio = true; 716 if (!strcmp(optarg, "native")) { 717 flags |= BDRV_O_NATIVE_AIO; 718 } else if (!strcmp(optarg, "threads")) { 719 /* this is the default */ 720 } else { 721 error_report("invalid aio mode `%s'", optarg); 722 exit(EXIT_FAILURE); 723 } 724 break; 725 case QEMU_NBD_OPT_DISCARD: 726 if (seen_discard) { 727 error_report("--discard can only be specified once"); 728 exit(EXIT_FAILURE); 729 } 730 seen_discard = true; 731 if (bdrv_parse_discard_flags(optarg, &flags) == -1) { 732 error_report("Invalid discard mode `%s'", optarg); 733 exit(EXIT_FAILURE); 734 } 735 break; 736 case QEMU_NBD_OPT_DETECT_ZEROES: 737 detect_zeroes = 738 qapi_enum_parse(&BlockdevDetectZeroesOptions_lookup, 739 optarg, 740 BLOCKDEV_DETECT_ZEROES_OPTIONS_OFF, 741 &local_err); 742 if (local_err) { 743 error_reportf_err(local_err, 744 "Failed to parse detect_zeroes mode: "); 745 exit(EXIT_FAILURE); 746 } 747 if (detect_zeroes == BLOCKDEV_DETECT_ZEROES_OPTIONS_UNMAP && 748 !(flags & BDRV_O_UNMAP)) { 749 error_report("setting detect-zeroes to unmap is not allowed " 750 "without setting discard operation to unmap"); 751 exit(EXIT_FAILURE); 752 } 753 break; 754 case 'b': 755 bindto = optarg; 756 break; 757 case 'p': 758 port = optarg; 759 break; 760 case 'o': 761 if (qemu_strtou64(optarg, NULL, 0, &dev_offset) < 0) { 762 error_report("Invalid offset '%s'", optarg); 763 exit(EXIT_FAILURE); 764 } 765 break; 766 case 'l': 767 if (strstart(optarg, SNAPSHOT_OPT_BASE, NULL)) { 768 sn_opts = qemu_opts_parse_noisily(&internal_snapshot_opts, 769 optarg, false); 770 if (!sn_opts) { 771 error_report("Failed in parsing snapshot param `%s'", 772 optarg); 773 exit(EXIT_FAILURE); 774 } 775 } else { 776 sn_id_or_name = optarg; 777 } 778 /* fall through */ 779 case 'r': 780 nbdflags |= NBD_FLAG_READ_ONLY; 781 flags &= ~BDRV_O_RDWR; 782 break; 783 case 'P': 784 warn_report("The '-P' option is deprecated; use --image-opts with " 785 "a raw device wrapper for subset exports instead"); 786 if (qemu_strtoi(optarg, NULL, 0, &partition) < 0 || 787 partition < 1 || partition > 8) { 788 error_report("Invalid partition '%s'", optarg); 789 exit(EXIT_FAILURE); 790 } 791 break; 792 case 'B': 793 bitmap = optarg; 794 break; 795 case 'k': 796 sockpath = optarg; 797 if (sockpath[0] != '/') { 798 error_report("socket path must be absolute"); 799 exit(EXIT_FAILURE); 800 } 801 break; 802 case 'd': 803 disconnect = true; 804 break; 805 case 'c': 806 device = optarg; 807 break; 808 case 'e': 809 if (qemu_strtoi(optarg, NULL, 0, &shared) < 0 || 810 shared < 1) { 811 error_report("Invalid shared device number '%s'", optarg); 812 exit(EXIT_FAILURE); 813 } 814 break; 815 case 'f': 816 fmt = optarg; 817 break; 818 case 't': 819 persistent = 1; 820 break; 821 case 'x': 822 export_name = optarg; 823 break; 824 case 'D': 825 export_description = optarg; 826 break; 827 case 'v': 828 verbose = 1; 829 break; 830 case 'V': 831 version(argv[0]); 832 exit(0); 833 break; 834 case 'h': 835 usage(argv[0]); 836 exit(0); 837 break; 838 case '?': 839 error_report("Try `%s --help' for more information.", argv[0]); 840 exit(EXIT_FAILURE); 841 case QEMU_NBD_OPT_OBJECT: { 842 QemuOpts *opts; 843 opts = qemu_opts_parse_noisily(&qemu_object_opts, 844 optarg, true); 845 if (!opts) { 846 exit(EXIT_FAILURE); 847 } 848 } break; 849 case QEMU_NBD_OPT_TLSCREDS: 850 tlscredsid = optarg; 851 break; 852 case QEMU_NBD_OPT_IMAGE_OPTS: 853 imageOpts = true; 854 break; 855 case 'T': 856 g_free(trace_file); 857 trace_file = trace_opt_parse(optarg); 858 break; 859 case QEMU_NBD_OPT_TLSAUTHZ: 860 tlsauthz = optarg; 861 break; 862 case QEMU_NBD_OPT_FORK: 863 fork_process = true; 864 break; 865 case 'L': 866 list = true; 867 break; 868 } 869 } 870 871 if (list) { 872 if (argc != optind) { 873 error_report("List mode is incompatible with a file name"); 874 exit(EXIT_FAILURE); 875 } 876 if (export_name || export_description || dev_offset || partition || 877 device || disconnect || fmt || sn_id_or_name || bitmap || 878 seen_aio || seen_discard || seen_cache) { 879 error_report("List mode is incompatible with per-device settings"); 880 exit(EXIT_FAILURE); 881 } 882 if (fork_process) { 883 error_report("List mode is incompatible with forking"); 884 exit(EXIT_FAILURE); 885 } 886 } else if ((argc - optind) != 1) { 887 error_report("Invalid number of arguments"); 888 error_printf("Try `%s --help' for more information.\n", argv[0]); 889 exit(EXIT_FAILURE); 890 } else if (!export_name) { 891 export_name = ""; 892 } 893 894 qemu_opts_foreach(&qemu_object_opts, 895 user_creatable_add_opts_foreach, 896 NULL, &error_fatal); 897 898 if (!trace_init_backends()) { 899 exit(1); 900 } 901 trace_init_file(trace_file); 902 qemu_set_log(LOG_TRACE); 903 904 socket_activation = check_socket_activation(); 905 if (socket_activation == 0) { 906 setup_address_and_port(&bindto, &port); 907 } else { 908 /* Using socket activation - check user didn't use -p etc. */ 909 const char *err_msg = socket_activation_validate_opts(device, sockpath, 910 bindto, port, 911 list); 912 if (err_msg != NULL) { 913 error_report("%s", err_msg); 914 exit(EXIT_FAILURE); 915 } 916 917 /* qemu-nbd can only listen on a single socket. */ 918 if (socket_activation > 1) { 919 error_report("qemu-nbd does not support socket activation with %s > 1", 920 "LISTEN_FDS"); 921 exit(EXIT_FAILURE); 922 } 923 } 924 925 if (tlscredsid) { 926 if (sockpath) { 927 error_report("TLS is only supported with IPv4/IPv6"); 928 exit(EXIT_FAILURE); 929 } 930 if (device) { 931 error_report("TLS is not supported with a host device"); 932 exit(EXIT_FAILURE); 933 } 934 if (tlsauthz && list) { 935 error_report("TLS authorization is incompatible with export list"); 936 exit(EXIT_FAILURE); 937 } 938 tlscreds = nbd_get_tls_creds(tlscredsid, list, &local_err); 939 if (local_err) { 940 error_report("Failed to get TLS creds %s", 941 error_get_pretty(local_err)); 942 exit(EXIT_FAILURE); 943 } 944 } else { 945 if (tlsauthz) { 946 error_report("--tls-authz is not permitted without --tls-creds"); 947 exit(EXIT_FAILURE); 948 } 949 } 950 951 if (list) { 952 saddr = nbd_build_socket_address(sockpath, bindto, port); 953 return qemu_nbd_client_list(saddr, tlscreds, bindto); 954 } 955 956 #if !HAVE_NBD_DEVICE 957 if (disconnect || device) { 958 error_report("Kernel /dev/nbdN support not available"); 959 exit(EXIT_FAILURE); 960 } 961 #else /* HAVE_NBD_DEVICE */ 962 if (disconnect) { 963 int nbdfd = open(argv[optind], O_RDWR); 964 if (nbdfd < 0) { 965 error_report("Cannot open %s: %s", argv[optind], 966 strerror(errno)); 967 exit(EXIT_FAILURE); 968 } 969 nbd_disconnect(nbdfd); 970 971 close(nbdfd); 972 973 printf("%s disconnected\n", argv[optind]); 974 975 return 0; 976 } 977 #endif 978 979 if ((device && !verbose) || fork_process) { 980 int stderr_fd[2]; 981 pid_t pid; 982 int ret; 983 984 if (qemu_pipe(stderr_fd) < 0) { 985 error_report("Error setting up communication pipe: %s", 986 strerror(errno)); 987 exit(EXIT_FAILURE); 988 } 989 990 /* Now daemonize, but keep a communication channel open to 991 * print errors and exit with the proper status code. 992 */ 993 pid = fork(); 994 if (pid < 0) { 995 error_report("Failed to fork: %s", strerror(errno)); 996 exit(EXIT_FAILURE); 997 } else if (pid == 0) { 998 close(stderr_fd[0]); 999 ret = qemu_daemon(1, 0); 1000 1001 /* Temporarily redirect stderr to the parent's pipe... */ 1002 old_stderr = dup(STDERR_FILENO); 1003 dup2(stderr_fd[1], STDERR_FILENO); 1004 if (ret < 0) { 1005 error_report("Failed to daemonize: %s", strerror(errno)); 1006 exit(EXIT_FAILURE); 1007 } 1008 1009 /* ... close the descriptor we inherited and go on. */ 1010 close(stderr_fd[1]); 1011 } else { 1012 bool errors = false; 1013 char *buf; 1014 1015 /* In the parent. Print error messages from the child until 1016 * it closes the pipe. 1017 */ 1018 close(stderr_fd[1]); 1019 buf = g_malloc(1024); 1020 while ((ret = read(stderr_fd[0], buf, 1024)) > 0) { 1021 errors = true; 1022 ret = qemu_write_full(STDERR_FILENO, buf, ret); 1023 if (ret < 0) { 1024 exit(EXIT_FAILURE); 1025 } 1026 } 1027 if (ret < 0) { 1028 error_report("Cannot read from daemon: %s", 1029 strerror(errno)); 1030 exit(EXIT_FAILURE); 1031 } 1032 1033 /* Usually the daemon should not print any message. 1034 * Exit with zero status in that case. 1035 */ 1036 exit(errors); 1037 } 1038 } 1039 1040 if (device != NULL && sockpath == NULL) { 1041 sockpath = g_malloc(128); 1042 snprintf(sockpath, 128, SOCKET_PATH, basename(device)); 1043 } 1044 1045 server = qio_net_listener_new(); 1046 if (socket_activation == 0) { 1047 saddr = nbd_build_socket_address(sockpath, bindto, port); 1048 if (qio_net_listener_open_sync(server, saddr, &local_err) < 0) { 1049 object_unref(OBJECT(server)); 1050 error_report_err(local_err); 1051 exit(EXIT_FAILURE); 1052 } 1053 } else { 1054 size_t i; 1055 /* See comment in check_socket_activation above. */ 1056 for (i = 0; i < socket_activation; i++) { 1057 QIOChannelSocket *sioc; 1058 sioc = qio_channel_socket_new_fd(FIRST_SOCKET_ACTIVATION_FD + i, 1059 &local_err); 1060 if (sioc == NULL) { 1061 object_unref(OBJECT(server)); 1062 error_report("Failed to use socket activation: %s", 1063 error_get_pretty(local_err)); 1064 exit(EXIT_FAILURE); 1065 } 1066 qio_net_listener_add(server, sioc); 1067 object_unref(OBJECT(sioc)); 1068 } 1069 } 1070 1071 if (qemu_init_main_loop(&local_err)) { 1072 error_report_err(local_err); 1073 exit(EXIT_FAILURE); 1074 } 1075 bdrv_init(); 1076 atexit(qemu_nbd_shutdown); 1077 1078 srcpath = argv[optind]; 1079 if (imageOpts) { 1080 QemuOpts *opts; 1081 if (fmt) { 1082 error_report("--image-opts and -f are mutually exclusive"); 1083 exit(EXIT_FAILURE); 1084 } 1085 opts = qemu_opts_parse_noisily(&file_opts, srcpath, true); 1086 if (!opts) { 1087 qemu_opts_reset(&file_opts); 1088 exit(EXIT_FAILURE); 1089 } 1090 options = qemu_opts_to_qdict(opts, NULL); 1091 qemu_opts_reset(&file_opts); 1092 blk = blk_new_open(NULL, NULL, options, flags, &local_err); 1093 } else { 1094 if (fmt) { 1095 options = qdict_new(); 1096 qdict_put_str(options, "driver", fmt); 1097 } 1098 blk = blk_new_open(srcpath, NULL, options, flags, &local_err); 1099 } 1100 1101 if (!blk) { 1102 error_reportf_err(local_err, "Failed to blk_new_open '%s': ", 1103 argv[optind]); 1104 exit(EXIT_FAILURE); 1105 } 1106 bs = blk_bs(blk); 1107 1108 blk_set_enable_write_cache(blk, !writethrough); 1109 1110 if (sn_opts) { 1111 ret = bdrv_snapshot_load_tmp(bs, 1112 qemu_opt_get(sn_opts, SNAPSHOT_OPT_ID), 1113 qemu_opt_get(sn_opts, SNAPSHOT_OPT_NAME), 1114 &local_err); 1115 } else if (sn_id_or_name) { 1116 ret = bdrv_snapshot_load_tmp_by_id_or_name(bs, sn_id_or_name, 1117 &local_err); 1118 } 1119 if (ret < 0) { 1120 error_reportf_err(local_err, "Failed to load snapshot: "); 1121 exit(EXIT_FAILURE); 1122 } 1123 1124 bs->detect_zeroes = detect_zeroes; 1125 fd_size = blk_getlength(blk); 1126 if (fd_size < 0) { 1127 error_report("Failed to determine the image length: %s", 1128 strerror(-fd_size)); 1129 exit(EXIT_FAILURE); 1130 } 1131 1132 if (dev_offset >= fd_size) { 1133 error_report("Offset (%" PRIu64 ") has to be smaller than the image " 1134 "size (%" PRId64 ")", dev_offset, fd_size); 1135 exit(EXIT_FAILURE); 1136 } 1137 fd_size -= dev_offset; 1138 1139 if (partition) { 1140 uint64_t limit; 1141 1142 if (dev_offset) { 1143 error_report("Cannot request partition and offset together"); 1144 exit(EXIT_FAILURE); 1145 } 1146 ret = find_partition(blk, partition, &dev_offset, &limit); 1147 if (ret < 0) { 1148 error_report("Could not find partition %d: %s", partition, 1149 strerror(-ret)); 1150 exit(EXIT_FAILURE); 1151 } 1152 /* 1153 * MBR partition limits are (32-bit << 9); this assert lets 1154 * the compiler know that we can't overflow 64 bits. 1155 */ 1156 assert(dev_offset + limit >= dev_offset); 1157 if (dev_offset + limit > fd_size) { 1158 error_report("Discovered partition %d at offset %" PRIu64 1159 " size %" PRIu64 ", but size exceeds file length %" 1160 PRId64, partition, dev_offset, limit, fd_size); 1161 exit(EXIT_FAILURE); 1162 } 1163 fd_size = limit; 1164 } 1165 1166 export = nbd_export_new(bs, dev_offset, fd_size, export_name, 1167 export_description, bitmap, nbdflags, 1168 nbd_export_closed, writethrough, NULL, 1169 &error_fatal); 1170 1171 if (device) { 1172 #if HAVE_NBD_DEVICE 1173 int ret; 1174 1175 ret = pthread_create(&client_thread, NULL, nbd_client_thread, device); 1176 if (ret != 0) { 1177 error_report("Failed to create client thread: %s", strerror(ret)); 1178 exit(EXIT_FAILURE); 1179 } 1180 #endif 1181 } else { 1182 /* Shut up GCC warnings. */ 1183 memset(&client_thread, 0, sizeof(client_thread)); 1184 } 1185 1186 nbd_update_server_watch(); 1187 1188 /* now when the initialization is (almost) complete, chdir("/") 1189 * to free any busy filesystems */ 1190 if (chdir("/") < 0) { 1191 error_report("Could not chdir to root directory: %s", 1192 strerror(errno)); 1193 exit(EXIT_FAILURE); 1194 } 1195 1196 if (fork_process) { 1197 dup2(old_stderr, STDERR_FILENO); 1198 close(old_stderr); 1199 } 1200 1201 state = RUNNING; 1202 do { 1203 main_loop_wait(false); 1204 if (state == TERMINATE) { 1205 state = TERMINATING; 1206 nbd_export_close(export); 1207 nbd_export_put(export); 1208 export = NULL; 1209 } 1210 } while (state != TERMINATED); 1211 1212 blk_unref(blk); 1213 if (sockpath) { 1214 unlink(sockpath); 1215 } 1216 1217 qemu_opts_del(sn_opts); 1218 1219 if (device) { 1220 void *ret; 1221 pthread_join(client_thread, &ret); 1222 exit(ret != NULL); 1223 } else { 1224 exit(EXIT_SUCCESS); 1225 } 1226 } 1227